We'd like to open up asynchronous copies in Archive:+copy-packages to everyone in order to be able to close several criticals and remove a pile of old code. However, there's some concern that Ubuntu auto-sync runs, which create thousands of PackageCopyJobs, will starve interactive copies between PPAs. Due to various production limitations there appears to be no straightforward way for us to scale up our PCJ-running capacity right now. In any case, the problem is not really one of capacity (carob only has relevant logs since 20121004 right now, but the peak since then has been 256 copies per day, 126 of which were to the Ubuntu primary archive; the longest the queue has been is 49, and that took 23 seconds to process) but of prioritisation.
I don't think the flow rate of routine copy jobs is sufficient to justify any work on prioritising between those, but Ubuntu auto-syncs are not routine: the first one of a cycle usually involves the submission of a couple of thousand PCJs in a matter of minutes (not quite instant because Archive.copyPackages times out with an excessively long list, so we have to subdivide it). We really don't care about how long it takes to process those copies, since it's going to take them days to build anyway, and it would be entirely fine for them to be preempted by other copies.
Auto-syncs are one of the very few users of Archive.copyPackages, which creates PCJs with copy_policy MASS_SYNC. Therefore, a reasonable prioritisation policy would seem to be to prioritise MASS_SYNC behind everything else.
It's hard to get accurate data automatically since there's nothing persistent that delimits one Archive.